Download Making the Numbers? “Short Termism” & the Puzzle of Only Occasional Disaster

Making the Numbers?
“Short Termism” & the
Puzzle of Only Occasional
Disaster
Hazhir Rahmandad
Nelson P Repenning
Rebecca Henderson
Working Paper
15-027
October 17, 2014
Copyright © 2014 by Hazhir Rahmandad, Nelson P Repenning, and Rebecca Henderson
Working papers are in draft form. This working paper is distributed for purposes of comment and
discussion only. It may not be reproduced without permission of the copyright holder. Copies of working
papers are available from the author.
MAKING THE NUMBERS?
“SHORT TERMISM” & THE PUZZLE OF ONLY OCCASIONAL DISASTER
Hazhir Rahmandad
Visiting Associate Professor, MIT Sloan
Associate Professor of Industrial and Systems Engineering, Virginia Tech
Nelson P Repenning
School of Management Distinguished Professor of System Dynamics and Organizations Studies
MIT
<[email protected]>
Rebecca Henderson
John and Natty McArthur University Professor, Harvard University and Faculty Research
Fellow, NBER
1
Abstract
Much recent work in strategy and popular discussion suggests that an excessive focus on “managing the
numbers” --delivering quarterly earnings at the expense of longer term investments--makes it difficult for
firms to make the investments necessary to build competitive advantage. “Short termism” has been
blamed for everything from the decline of the US automobile industry to the low penetration of
techniques such as TQM and continuous improvement. Yet a vigorous tradition in the accounting
literature establishes that firms routinely sacrifice long-term investment to manage earnings and are
rewarded for doing so. This paper presents a model that reconciles these apparently contradictory
perspectives. We show that if the source of long-term advantage is modeled as a stock of capability that
accumulates over time, a firm’s proclivity to manage short-term earnings at the expense of long-term
investment can have very different consequences depending on whether the firm’s capability is close to a
critical “tipping threshold”. When the firm operates above this threshold, managing earnings smoothes
revenue and cash flow with few long-term consequences. Below it, managing earnings can tip the firm
into a vicious cycle of accelerating decline. Our results have important implications for understanding
managerial incentives and the internal processes that create sustained advantage.
1
Everyone who has worked with American managements can testify that the need to satisfy the pension
fund manager's quest for higher earnings next quarter, together with the panicky fear of the raider,
constantly pushes top managements toward decisions they know to be costly, if not suicidal, mistakes. The
damage is greatest where we can least afford it: in the fast-growing middle-sized high-tech or highengineering firm that needs to put every available penny into tomorrow. (Drucker 1986) (Quoted in
(Laverty 1996))
1. Introduction.
There is little doubt that the desire for smooth, reliable earnings has a significant effect on
managerial behavior. Even the most casual visit to a publicly-traded company during a period of lowerthan-expected revenue reveals the lengths to which managers will go to “meet their numbers”, doing
anything from banning travel and eliminating training to canceling investment projects and delaying
maintenance. Indeed, research within the finance and accounting literatures finds that managers do
sacrifice (at least some) long-term investments in response to pressure from the capital markets. Graham
et al.’s, (2005) large-scale qualitative study of CFOs and CEOs finds that 78% of the managers surveyed
admit to forgoing some long-term value in favor of smoother earnings. Healy and Wahlan (1999), in a
summary of the literature suggest that earnings management is pervasive, while Brav et al. (2005) find
that capital market pressures lead managers to “avoid cutting dividends at all costs” even if this means
bypassing NPV positive investments and Bartov (1993) shows that managers routinely time asset sales to
smooth inter-temporal earnings changes. Benner (2007; 2010) has further suggested that firms going
through significant technological transitions face particularly intense pressure from analysts, causing them
to reduce capital investment and investment in R&D. But does such a focus on short-term results actually
detract from overall firm performance? Despite the ubiquity of this practice, both popular opinion and the
available data remain decidedly mixed and the strategy literature takes a distinctly different perspective
from that which dominates the finance and accounting literatures.
Research in strategy has long stressed the importance of sustained investment over long periods
of time as a key source of competitive advantage. Scholars working within this tradition have suggested
that the unique resources and distinctive organizational competences that underlie enduring supra-normal
1
returns cannot be developed instantaneously but must be “grown” gradually over time (Dierickx and Cool
1989). For example, Wal-Mart’s superior IT capabilities were built through sustained investment over
many years; by the time K-Mart attempted to respond by building equivalent skills the investment
required was more than their balance sheet could sustain. Similarly the organizational competencies
developed by firms such as Nordstrom, Southwest, Toyota, McKinsey and Merck appear to rest on the
slow accumulation of mutual trust and local or “sticky” knowledge that takes years to develop (Helfat
2007; Gibbons and Henderson 2010).
This understanding has led to the suggestion that the short term orientation of American capital
markets and American managers is a significant problem. Commenting on the recent meltdown of the US
financial system, William Donaldson, a former head of the securities and exchange commission, said
“The excessive focus by too many corporations on achieving short-term results… certainly is one of the
root causes for some of the problems we face today (Rummel 2008).” Donaldson is not alone in his
views: both academics and pundits are fond of highlighting the “tyranny of quarterly earnings” as the
principal reason for the perceived decline of the western economy (Laverty 1996; The Aspen Institute
2009; Haldane 2010; Dallas 2011).
Consistent with this view, there are plenty of examples of once successful companies whose
efforts to achieve short-term performance targets appear to have had a negative impact on their
willingness to make important longer term investments. GM’s recent recall challenges are thought to
stem from short-sighted decisions and, more generally, the ongoing woes of the US auto industry are
often attributed to their unwillingness to address the eventual need for smaller fuel-efficient vehicles in
favor of continuing to sell profitable trucks and sport utility vehicles (e.g., Ingrassia 2010). Similarly,
Pfeffer and Sutton (2000: 142-147) document how a focus on short-term results severely weakened
Hewlett-Packard’s much vaunted employee focused culture, while the explosion at BP’s Texas City plant
that killed 15 people was blamed on BP’s short term orientation (Baker, Bowman et al. 2007). At a more
micro level, a sequence of papers exploring the dynamics of process-focused improvement suggests that
2
many firms are reluctant to invest in practices that raise profitability in the long term if they come at the
expense of short term performance (Easton and Jarrell 1998; Repenning and Sterman 2001; Repenning
and Sterman 2002). Work in the innovation tradition suggests that one of the reasons firms experience
difficulty in responding to discontinuous technological change is that they are reluctant to make
investments in the next generation technologies at the expense of investment in the status quo
(Christensen 1997). Similarly, work in organizational economics suggests that a focus on short-term
measureable performance can induce managers to neglect those tasks for which clear performance
measures are not available (Holmstrom and Milgrom 1991), and that rewarding short-term performance
when managerial ability is unclear can lead managers to “borrow” earnings from the future at an
unfavorable rate (Stein 1989).
In contrast to this collection of case evidence and theory most large sample studies in the
accounting literature suggest that “managing earnings” and “meeting analyst expectations” are
systematically associated with superior performance. While Bhojraj et al. (2009) find that firms that
manage to marginally beat earnings forecast through manipulation improve their stock price in the short
term but have lower ROA over time, Barth et al. (1999) find that firms that report continuous growth in
annual earnings earn a premium in the market. Similarly, Bartov et al. (2002) find that firms that meet or
beat analyst expectations often report superior operating profits and Gunny (2010) finds that taking “real”
actions such as reducing R&D or reducing SG&A to increase income has a positive effect on subsequent
ROA.
A significant body of theory explains these results by suggesting that they reflect agency or
information problems. Managers who “make their numbers” may be signaling both that they know what
they are doing and that the firm’s performance is under control. For example, Bowen et al. (1995) and
Burgstahler and Dicheu (1997) suggest that those who “manage earnings” enhance their reputation with
key stakeholders and thus obtain better terms of trade while Farrell and Whichbee (2003) suggest that
3
“making the numbers” reflects career concerns on the part of senior managers who would otherwise lose
their jobs.
The idea that sacrificing some long-term investment for the sake of short-term earnings
management might not be entirely disastrous is further supported by the observation that, to paraphrase
Mark Twain, reports of the demise of the western economy have often been greatly exaggerated. A
former SEC chief, Harold Williams, said, in an address given in November 1980 (Williams 1980), “I
believe that the most fundamental problem facing us is that we have lost our sense of the future—we tend
increasingly to focus on the short-run and to ignore the longer-range consequences of business and
political decisions.” Williams’ remarks failed to predict what would eventually become the longest
peacetime expansion in US economic history, and there is little reason to believe that managers became
less shortsighted during the period. Similarly while there is general agreement that the managers of
publicly traded companies in Germany and Japan face significantly less pressure from their investors for
short term performance, there is no systematic evidence that they outperform their UK or US listed rivals
or that the German or Japanese economies are any more innovative than those of the US (Peek and
Rosengren 2005; Dustmann, Fitzenberger et al. 2014).
How are we to make sense of this apparent paradox? When does a short term, target-driven focus
increase performance and when does it lead to the erosion of critical long-term capabilities and a failure
to invest in new markets and new technologies? In this paper we attempt to answer this question through
the development of a simple model that captures the dynamics associated with managing earnings. Our
formulation builds on a premise central to the strategy literature: long-term investment leads to the
development of a stock of “capability” which continually erodes through natural entropy and thus requires
continued investment and maintenance. We think of capability as including many competitively
significant assets – particularly organizational capabilities, but also such intangibles as reputation and
brand value. Critically, the stock of capability and its effect on performance can neither be easily
measured nor predicted by the firm. Investments in capability are thus likely to be driven more by rules of
4
thumb and managerial intuition than by precise calculation. We then define firm performance as a
function of the effort devoted to short-term performance and the current level of capability, thus
presenting the firm with the task of trading off investments in long-term capability against short-term
revenue generation. We use the model to study the dynamics that emerge when firms try to “manage
earnings” by changing the emphasis placed on short-run revenue performance in response to exogenous
shocks to revenue.
Analyzing our model yields several insights. First, we show that conceptualizing capability as a
stock or level variable implies that the firm faces a temporal trade-off in making resource allocation
decisions. If managerial time is scarce, increasing the attention dedicated to capabilities necessarily
results in a temporary decline in overall performance – a pattern of “worse before better”. Conversely,
reducing the time dedicated to cultivating capabilities leads to a temporary increase in performance – a
pattern of “better before worse.” Second, combining these physical dynamics with the efforts of
managers to smooth earnings yields a system characterized by two stable equilibria separated by a tipping
threshold. Critically, this structure implies that attempts to manage earnings can produce very different
outcomes depending on their intensity. As long as the resource adjustments targeted at smoothing
performance keep the system above its tipping threshold, they will have a only a second-order effect on
the long-run average performance while significantly reducing its variation around the mean. Above the
tipping threshold managing earnings thus offers several positive benefits including improved internal
financial discipline and higher external investor confidence. Since it seems plausible that this is the
prevailing state most of the time in most industries, our analysis suggests that it is not surprising that the
accounting literature finds that most firms smooth earnings and are rewarded by the market for doing so.
However, if, in an effort to maintain earnings stability, capability is allowed to fall below this critical
threshold, then efforts to smooth earnings become a path to failure. The firm enters a vicious spiral
created by an increased emphasis on revenue generation and declining organizational capability which,
left unchecked, eventually leads to failure.
5
Third, we use the model to identify the conditions under which firms are most prone to “tipping”
into this pathological set of dynamics. We show that the risk of tipping into a vicious spiral of declining
capability increases with the manager’s willingness to shift investment in response to short term revenue
shortfalls. Those managers who “overweight” short-term returns even when times are good are
significantly more likely to tip towards disaster.
Finally, we explore the effects of increasing demand variability on these dynamics. We show that
when demand is relatively smooth, firms that rapidly shift investment to keep day to day revenues stable
minimize revenue variability and thus presumably maximize their equity performance with no significant
long-term penalty. However as demand variability increases, or lasts for longer periods of time they
become significantly more likely to tip into disaster than the firms that respond less dramatically to
shortfalls in immediate revenue.
These results have several implications for our understanding of the sources of competitive
advantage. Most critically, they suggest that whether earnings management is a source of improved
performance or deeply pathological is a matter of degree, and that the dynamics of the system are such
that it is relatively easy for managers to learn the wrong lessons about the optimal mix of the two
activities. Unless one holds to heroic assumptions about managers’ understanding of the system in which
they are embedded, it appears that managers cannot necessarily be counted on to make this trade-off in
the best interests of the firm over the long-term. Our results thus suggest that organizational practices that
protect investments in capability from the inevitable desire to cut them during down periods are likely to
be a key source of persistent above average firm performance. Most notably, compensation systems that
rely on relational contracts and subjective performance measures to reward the long term development of
capability may prove to be a key contributor to sustained above-average performance.
We organize our arguments as follows. In the next section we present our model, first analyzing
the implications of our conceptualization of capability and performance. Following that, we add a
6
decision rule to the model in which managers treat earnings management as a simple control problem and
then analyze its dynamics. Finally, we discuss the implications of our results for future research.
2.0 The Model
2.1 Capability and Performance
We begin the specification of our model with the notion of capability. Following Dierickx and
Cool (1989) we hypothesize that because they cannot be acquired directly, capabilities are usefully
conceptualized as being the outcome of a process of accumulation and therefore behave as stock or level
variables. Managers in this formulation cannot influence the level of capability directly, but instead
control only its rate of change. Thus, for example, managers cannot immediately transform a poorly
maintained production line and a poorly trained workforce into a high performance operation. Instead,
they must invest in maintenance, process improvement and training, and, with time, operational
performance will improve. Conversely, if managers suddenly begin to neglect maintenance and training,
operational performance does not immediately decline; time is required before poorly maintained
equipment begins to break and the combination of turnover and forgetting yield a poorly skilled
workforce.
Consistent with prior models of capability dynamics (Rahmandad 2012), we model capability, C,
as a stock or level variable whose rate of change is determined by two factors, investment and erosion:
(1)
Investment, the first term, is the product of the effort the firm dedicates to developing and
maintaining capability, ec, and the productivity of those efforts, . Erosion is captured by the second
term: absent investment, capability erodes with an average delay of 1/.1 We specify the firm’s revenue,
1
. Note our formulation is simpler than the one proposed by Dierickx and Cool (1989) who suggest that managers
may also influence the rate with which capability erodes. For simplicity, we assume that erosion happens
autonomously and that managers influence capability solely through the investment term.
7
R, as a constant return to scale Cobb-Douglas function of the firm’s capability, C, the effort currently
being dedicated to generating revenue, eR, and an environmental shock, S:
(2)
Finally, we assume that the resources available to the firm are bounded by h, the number of
available hours of effort, and define u as the fraction of that effort allocated to revenue generation.
(3)
(4)
In the long run firms could in principle choose to increase h but to focus attention on the impact
of earning management we assume h is fixed. Calculating the steady state optimal allocation fraction, u*,
which assumes a constant allocation fraction and steady-state capability level, is straightforward:
(5)
2.2 Earnings Management and Capability Investment
The evidence suggests considerable variation in how different firms trade-off mean performance and
variability around that mean. To study how such differences in preferences might influence the system’s
dynamics, we use an investment allocation rule that allows us to easily vary the firm’s focus across a
continuum ranging from the short to the long run. On one side of that spectrum we imagine a firm that
cares about meeting short-term revenue targets and uses the following decision rule:
(
(
) )
(6.1)
Our hypothesized managers begin with the steady-state optimal allocation, u*, and then adjust that
allocation to try to minimize the gap between the current revenue under u*, R(u*,S,C), and the market’s
expectation. Markets likely do not observe current capabilities directly, so market expectation,
R(u*,0,C*), reflects a firm that is well managed, i.e. has a capability around the optimal (C*) level.
Moreover, the market does not foresee environmental shocks or discounts them due to agency concerns
(e.g., Stein 1989). The parameter γ represents the intensity of the firm’s effort to offset environmental
8
shocks to maintain smooth earnings and varies between 0 and 1/β, with =1/ representing a situation in
which the firm attempts to completely offset any revenue excess or shortfall by reallocating effort.
On the other end of the continuum, we imagine a firm that cares only about expected long term
results and acts rationally. Solving the finite horizon dynamic program yields (approximately) the
following decision rule for our long run (LR) firm (see appendix for details)2:
(
( ))
(6.2)
The firm focused on the long run pays no attention to earnings, which are subject to environmental
shocks, but instead always works to insure that capability, C, is equal to its optimal level, C*. Combining
both the short and long decision rules (6.1 and 6.2) into a single equation yields the following decision
rule for :
(
(
) ( )
)
(6.3)
By adjusting  over [0,1/], our formulation allows us to analyze a continuum of decision rules, ranging
from an exclusive focus on meeting current expectations, =1/, to an exclusive focus on long run
performance, =0. Our formulation thus allows us to study two competing forces that shape managerial
action. The logic of a “rational” agent is captured with the capability term and the behavioral market
pressures to produce smooth performance are incorporated by the revenue adjustment term. By changing
γ we can examine the outcomes generated by any combination of long run “rational” choice and shortterm focus.
The remainder of the model’s specification follows the explanations above and is straightforward:
2
The solution to the finite horizon dynamic program includes a term for “end” effects (capturing the shift in
allocation to benefit from exhausting capability’s terminal value). To focus on the steady-state behavior (early in
the finite horizon) of managers and for simplicity we do not include that term here. The infinite horizon solution
(also discussed in the online appendix and briefly in the analysis) is very similar to equation 6.3, with an adjustment
for discounting, which does not change the results qualitatively.
9
(7)
(8)
(9)
(10)
2.3 Model Overview
These assumptions yield a system comprised of four key feedback loops (see figure 1). First, there is the
balancing Capability Erosion loop; absent ongoing investment, capability erodes towards zero, with an
average delay of . Second, the decision rule described above results in two balancing loops Make the
Numbers and Capability Preservation. As expected revenue, R(u*,S,C) falls below its target, R(u*,0,C*),
managers concerned with short term results (i.e. those with non zero values of  increase the effort
dedicated to revenue generation (due to term (
) in 6.3), which, in the short run, increases
revenue and closes the gap. Similarly, when capability deviates from its optimal value (C*), effort will be
allocated towards investing in it (due to term ( )
in 6.3). Finally, a reinforcing loop may emerge as a
result of the interaction between Make the Numbers3 and Capability Preservation loops. We call this the
Capability Tipping loop. Replacing equations 7 and 8 into 6.3 we can see that the Capability Tipping loop
is active only when γ>1, when a deviation in capability from the optimal value can lead to managerial
reactions that further reinforce that deviation. For γ<1 the effect of Capability Preservation loop on
capability levels is stronger than the Make the Numbers loop and thus any deviation from C* is balanced
in effort allocation. Therefore γ=1 is a critical threshold in 6.3, below which the Capability Preservation
mechanism dominates the dynamics and above which Capability Tipping becomes salient.
Notice that the Capability Tipping loop can act as either a virtuous or vicious cycle and will tend
to amplify whichever behavior is currently in progress. For example, exceeding the revenue target allows
3
“Make the Numbers” loop requires at least one state variable. Our formulations assume this loop adjusts quickly
relative to the dynamics of capability and thus the system can be reduced to 1 st order.
10
the firm to grow its stock of capability and makes achieving future revenue objectives even easier. When
operating in this virtuous direction the Capability Tipping loop creates a growing level of capability and
competitive advantage. If, however, revenue falls below its target, firms engaged in earnings
management may divert resources away from capability development in an effort to close the gap (the
Make the Numbers loop). While this will increase revenue in the short run, it also reduces investment in
capability. As the level of capability falls, it becomes more difficult for the firm to achieve its revenue
target, thus increasing the pressure to shift resources towards revenue generation and away from
capability development. In the extreme, capability can decline so much that the firm collapses –
reproducing the dynamics identified by those who lament the short-term focus of the capital markets. In
the next section, we use the model to explore the conditions under which this is more or less likely to
happen.
Average
Capability Life
B
Capability
Erosion
Capability (C)
+
Capability
Growth
(τ)
+ -
Capability
Erosion
+
Capability
Investment
Productivity
R
B
Capability
Tipping
Capability
Preservation
(ρ)
C
-
T
(R )
+
B
-
Effort to Develop
Capability
(e )
Target Revenue
+
Revenue
Revenue Effort
Fraction (u) +
Make the
Numbers
- +
Revenue Ratio
+
Managerial
Response (γ)
Figure 1-Model Overview
11
3. Model Analysis
3.1 Worse-Before-Better and Better-Before-Worse
We begin the analysis by reproducing two of the dynamics highlighted in the qualitative literature
on capability dynamics (e.g., Repenning and Sterman 2001): “Better before worse” and “Worse before
better”. Figure 2-a shows a set of model runs in which effort is initially split optimally (u=u*) between
revenue generation and capability investment and then, in month ten, is rebalanced to focus “excessively”
on generating short term revenue. We show three possible paths, each corresponding to a different
increment of a short term focus. In each case, immediately following the shift, because capability does
not immediately degrade, revenue temporarily grows above the steady state optimum. Later, however, as
the reduced investment causes capability to erode, revenue begins to fall. Adjusting effort towards
revenue generation thus creates a “better-before-worse” behavior pattern: by reducing investment in
capability the firm can temporarily raise its revenue, but such gains come at the expense of future
performance. This is the “milk the business” trajectory highlighted by those who worry about an
excessive focus on short-term performance.
Conversely, Figure 2-b shows the revenue trajectories when the simulation starts with varying
degrees of an overemphasis on revenue generation and, then, at month ten effort is rebalanced to match
the optimal mix, u*. In these cases revenue initially declines with the reduction in the effort dedicated to
generating revenue. Later, as the additional effort dedicated to capability investment begins to pay off,
revenue improves, eventually surpassing the initial level (since the effort balance is now optimal in steady
state). This is the “worse-before-better” trade-off found in the innovation and process improvement
literatures (e.g., Repenning and Sterman 2001): increased investment in assets and activities that build
capability pays off only after depressing short term results.
12
Figure 2- Better before worse and worse before better reference modes
As these results highlight, if investment in capability is usefully conceptualized as building assets that are
subject to the processes of accumulation and erosion, then managers face an important temporal trade-off.
Since the productive capabilities of these assets do not degrade immediately when investments in them
are reduced, managers have some ability to adjust when revenue is actually earned. A manager facing a
weak fourth quarter might, for example, choose to cut back on investment in training, maintenance or
R&D in favor of revenue generation to meet earnings targets, planning to rebuild the consequent loss in
capability later. Conversely, those who wish to push their organization to a higher level of performance
by developing new capabilities must endure a period of low performance before their investments yield
significant gains.
Note that these dynamics raise the agency concerns highlighted by Stein (1989) and others. A
principal (or a shareholder) observing a performance decline has difficulty determining whether this
decline is due to an increased investment in capability—which will pay off in the future—or to lack of
ability on the part of the agent (or the management team). Conversely, principals observing a
performance gain cannot determine whether it signals above average ability or simply results from the
agent focusing on short run gains at the expense of long run performance. Stein (1989) suggests that in
such cases managers will often “borrow” from the future at an unfavorable rate—akin to focusing on
13
revenue at the expense of capability-- in an effort to demonstrate that they are neither shirking nor
incompetent. Given their presumption of equilibrium, however, models in this tradition cannot analyze
the disequilibrium dynamics that may result from an overemphasis on immediate revenue generation,
particularly in the face of external shocks. In what follows we explore how managers’ efforts to maintain
steady earnings can have quite different effects depending on the state of the firm. While models like
Stein (1989) predict a uniform degradation in performance as a result of the pressure to maintain short
term results, our analysis shows that the influence of earnings management on performance can be nonlinear due to the existence of a critical threshold beyond which efforts to smooth earning shift from
functional to pathological.
3.2 Earnings Management and Tipping
To analyze the dynamics of earnings management, we begin with a simple case in which two
firms, one focused on delivering smooth earnings (=1.75) and one focused on long run performance
(=.25), are exposed to two offsetting revenue shocks, one unfavorable (at month 10) and one favorable
(at month 25). As shown in Figure 3-a, the firm focused on smoothing earnings responds to the negative
shock by reallocating effort away from capability development towards revenue generation, and,
consequently, the decline is revenue is modest and smaller than the shock that precipitated it. Capability
falls and stays below its optimal level (recovering only asymptotically) until the positive shock. When
the positive shock occurs, the firm allocates effort away from revenue generation and towards capability
development, thereby returning the system to its optimal operating point. The capability-focused firm, in
contrast, is less willing to divert resources away from investment and therefore experiences a larger
decline in revenue (see Figure 3-b). Capability degrades only slightly and recovers quickly. Conversely,
when the positive shock is experienced, the capability-focused firm passes the gains through to its owners
rather than building a reserve of capability.
Consider how the external analysts, who can only observe revenue directly, might evaluate the
two firms. Facing a sudden drop in capability (e.g. due to supply chain disruption; panels c and d) the
14
capability-focused manager increases allocation to capability, suffers a significant short-term reduction in
revenue, but is able to get back on track relatively quickly. In contrast the revenue-focused manager
increases the allocation to revenue and avoids the sharp drop in revenue, at the expense of slower
recovery of the capability. Both survive the exogenous shocks and generate similar amounts of
cumulative revenue. The capability-focused firm, however, is likely seen by market observers as offering
less reliable returns in comparison to its revenue-focused peer. In this example revenue smoothing has
few costs and goes a long way towards alleviating the agency concerns faced by owners who cannot
reliably observe capability or effort. We would thus expect that more revenue-focused heuristics (i.e.
higher values of γ in our setting) to be attractive for a notable fraction of managers.
Figure 3-Response to temporary revenue shocks
Though the revenue heuristic has several desirable properties, a more complete analysis suggests
that earnings management is not always so costless. To fully describe this system’s dynamics we use the
rate-level plot, a graph that relates the net rate of change in capability to the current level of capability
(Strogatz 1994). In our formulation capability change partially depends on environmental shocks. To
focus on the non-random structures however, we generate this plot by focusing on the deterministic
components of change in capability. Thus the equation for the rate-level plot can be found by replacing
equations 6.3 and 4 into 1, and simplifying the results:
15
̇
(
(
( )
))
(11)
The rate-level plot captures the evolution of firm’s state (capability) by showing how the net rate of
change in capability is a function of current capability level. The slope of the line indicates whether, in
that region, the system is dominated by a negative (downwardly sloping) or positive (upwardly sloping)
feedback loop. Intersections between the rate level plot and the x-axis represent equilibria (since
dC/dt=0) and the slope at the intersection determines the equilbrium’s stability. If the slope is positive,
the equilibrium is unstable—small perturbations will not be offset-- while a negative slope indicates an
equilibrium that is locally stable—small perturbations will be offset by the system’s dynamics.
The rate-level plot for the system we study can take one of two basic shapes depending on the
sign of γ-1 term in eq. 11. The first possibility is shown in Figure 4-a. If ≤1 (indicating an emphasis on
maintaining capability) then the plot is entirely downwardly sloping and intersects the x-axis at the
optimal capability level. The system thus has only one, stable, equilibrium and shows no possibility of
“tipping” or other more complex dynamics. From a feedback perspective, as long as ≤1 the capability
preservation loop is stronger than the Make the Numbers loop and, therefore, dominates the system’s
dynamics. As long as ≤1 the system’s dynamics will return it to the optimal capability level and there is
no possibility of failure.
If >1, however, the rate level plot has a more complex form (Figure 4-b). Now the phase plot
has three basic components, two downwardly sloping and one upwardly sloping, and crosses the x-axis
three times indicating the system has three equilibria. Starting at the right, the first equilibrium is stable
and occurs at the optimal capability level. In this region the balancing Capability Preservation feedback
process again dominates the system’s dynamics and, as indicated by the trajectory arrows, works to offset
any departures from the optimal capability level. As capability declines however, the phase plot becomes
positively sloped (between, in this case, capability levels of about 15 and 40) indicating a shift in
dominance to the reinforcing Capability Tipping loop. In the region in which the phase plot is both
16
upwardly sloping and lies above the axis, the Capability Tipping loop works in the upward (virtuous)
direction—more capability makes it easier to hit the revenue target, thus enabling further investment in
capability—and thus, as indicated on by the arrows on the phase plot, drives the system back to the target
capability level.
Figure 4-Sample Rate-Level plot for the System
Once the phase plot falls below the x axis, however, the dynamics change significantly. The
Capability Tipping feedback remains dominant but now works in the downward (vicious) direction—
declining capability makes it more difficult to achieve the revenue target thus necessitating a further shift
away from investment and towards revenue generation—and drives capability towards zero. The slope of
the plot once again becomes negative when capability has fallen to the point where all of the firm’s
resources are required to hit the revenue target (in this example at a capability level of about 15). At this
point investment in capability ceases and the negative Capability Erosion loop dominates the system’s
dynamics and drives capability to the second stable equilibrium at the origin.
The rate-level plot formalism makes the system’s dynamics easy to characterize. The equilibrium
at the right is stable and represents a firm at the optimal capability (C*) level. At this point, in steady
state, the resource allocation is optimal (u=u*) and thus delivers the target revenue, R(u*,0,C*), in
17
expectation. The local stability of this equilibrium implies that small perturbations will be offset by the
system’s dynamics (also indicated by the arrows pointing towards the equilibrium point). In other words,
while in the neighborhood of this point, the specified decision rule will work to offset any randomness in
revenue generation and thereby return the system to its optimal operating point. It is exactly this stability
that allows our hypothesized firm to smooth earnings with little penalty. Redirecting resources to revenue
generation (and thereby offsetting negatives shocks) causes capability to fall, but as long as capability
remains in the stable region it will eventually return to the operating point.
Moving to the far left of Figure 4-b, there is a second stable equilibrium at the origin. At this
point, investment has ceased, capability has fallen to zero, and essentially the firm has already failed. This
equilibrium is also locally stable; once in its neighborhood, absent an additional intervention, the firm will
be on a trajectory towards failure. Here, the revenue focus embodied in the decision rule (6.3), rather
than being functional, drives the firm to ruin; the efforts of managers to maintain revenue trap the firm in
a vicious cycle of declining capability and desperate efforts to boost revenue.
A third, unstable, equilibrium lies between the two stable ones. Due to its instability, even the
smallest perturbation will move the system towards one of the two stable points, thus making it unlikely
that the system would ever settle at this level of capability. Nonetheless, this point plays a critical role in
determining the system’s dynamics. As indicated by the arrows on the rate-level curve, it represents the
point at which the utility of Capability Tipping loop changes direction. When operating at levels of
capability above the unstable equilibrium, this feedback acts as a virtuous cycle and drives the system
back to the desired operating point. In this region, managers’ efforts to smooth earnings do not entail a
long-term sacrifice in performance and capability will tend to revert to the optimal level. Once capability
falls below the unstable equilibrium, however, the loop works as a vicious cycle—less capability, an
increased focus on revenue and declining investment-- and the same decision rule now drives the firm
towards failure. The unstable equilibrium thus represents a kind of “tipping point”: As long as capability
remains above it, the system remains stable and efforts to smooth earning can have considerable benefits
18
and few costs. If, however, efforts to smooth earnings push capability below that critical threshold, then,
absent an additional intervention or change in the decision rule, the firm will be on a trajectory towards
failure.
The main insight that thus emerges from our analysis is that whether efforts to smooth earnings
detract from performance is a matter of degree. As long as capability remains above the critical threshold
captured by the unstable equilibrium, efforts to adjust earning have only a second-order impact on
average performance. Once that threshold is crossed, however, the desire to smooth earnings become
pathological as the efforts to maintain revenue come at the expense of the firm’s long term viability. Of
course, our analysis presumes that managers who cross this threshold continue to use the decision rule
outlined in (6.3) and do not change their tendency to smooth earnings (represented by γ). This assumption
undoubtedly simplifies real behavior; it seems likely that many managers would eventually realize that
capability had fallen to a dangerously low level. If, however, managers only make this realization with a
significant delay, then a firm that crossed its capability threshold could find itself in a vicious cycle of
increasing emphasis on earning and declining investment in capability for some time before its leaders
recognized that their efforts to manage earnings were threatening the firm’s long term viability.
There are several reasons to believe that managers will not immediately recognize when their
efforts to manage earnings have moved from functional to pathological. Most generally, learning is
difficult in non-linear dynamic systems because local search and simple extrapolation are poor guides to
anticipating outcomes (Sterman 1994). Systems with tipping dynamics are particularly challenging on
this score because the system’s response to inputs when operating above the threshold differ from the
response once the threshold is crossed (see also Rudolph and Repenning 2002). Moreover managers who
do realize that they have crossed the tipping threshold face the “worse-before-better” dynamics associated
with an increased emphasis on capability that we described at the beginning of our analysis. Due to
agency concerns, managers who are rewarded principally on short-term results may not be able to
“afford” the price of rebalancing effort towards capability generation even if that was the right decision.
19
Figure 5 shows two model runs that illustrate the difficulty of learning to manage in this system.
In both cases a revenue-focused firm (γ=1.5) begins with the optimal level of capability and is exposed to
negative revenue shock, in the first case by 25%, and in the second by 30%.
Figure 5- Response to 25% and 30% Environmental Shocks to Revenue
In the case of a 25% shock (Figure 5-a), revenue suffers, but increasing the effort allocated to
revenue generation (Figure 5-c) partially offsets the reduction, keeping it below 25%. Revenue declines
during the “down period” as capability (Figure 5-d) begins to erode, but, once the shock is removed the
system returns to the target revenue and capability levels. The manager of this hypothetical firm would
likely be rewarded relative to her more capability-focused peers for exerting considerable “financial
discipline” during a difficult set of business conditions. The outcome is quite different, however, when
20
the environmental shock is slightly larger, in this case 30% instead of 25% ( Figure 5-b). Revenue
initially follows a path similar to that in 25% shock as revenue management kicks in (Figure 5-c).
However, the intensity of revenue management during the “down” year pushes the capability below the
tipping threshold. Having passed the tipping point, managers’ continued focus on revenue creates a
vicious cycle of declining earnings, increased effort to revenue generations, and the erosion of the
capability needed to generate those earnings (Figure 5-d). The reinforcing loop Capability Tipping thus
“takes over” and drives the firm towards failure.
Note how past experience does little to reveal the location of the tipping threshold. Imagine a
young manager in a cyclical industry charged with maintaining earnings through the inevitable down
periods. During her first down turn, she might make relatively modest reallocations (well above the
tipping threshold) of effort and the system would reward her for doing so. Based on that initial
experience, her reallocations are likely to intensify in each subsequent cycle and she will be increasingly
rewarded for doing so. Humans are, the word of psychologist Jim Reason, “furious pattern matchers”
(Reason 1990) and this pattern is all too easy to identify: the more aggressive her efforts, the better her
performance. Missing from this learning, however, is any sense that with each effort she is moving her
organization closer to the tipping threshold. Only when the decline in capability begins to manifest
(perhaps as an accident or a major product recall) will she have any sense of the full range of the system’s
dynamics. But even this experience can be attributed to a plethora of external causes and might need to
be repeated several times to temper what she learned when operating on the other side of the tipping
threshold.
Our analysis thus yields a candidate resolution to the puzzle of why managers’ efforts to “make
their numbers” can seem alternatively functional and pathological. Recognizing that capability behaves
as a stock or level variable combined with the efforts of managers to offset environmental shocks to
revenue generation creates a system with a tipping threshold. When operating above that threshold,
efforts to manage earnings are a functional response to the capital market’s desire for smooth earnings
21
and lead to both reduced agency concerns and higher valuations. Moreover, operating above the
threshold provides managers with powerful evidence concerning the utility of earnings management; the
more they do it, the more they are rewarded by the external markets. Once that threshold is crossed,
however, earnings management becomes a path to low performance. Whether or not the firm survives this
type of decline will depend on its strength relative to its competitors, its institutional context (e.g.,
whether a government is willing to support it through a rebuilding period) and, how long it takes
managers to recognize that a change of approach is in order.
3.3 Comparative Dynamics
The existence of a tipping threshold for γ>1 suggests that in the pursuit of stable earnings managers risk
descending from a higher performance configuration into a vicious cycle of declining performance and
eroding capability. As highlighted by the rate level plot the distance between the desired, stable,
equilibrium and the tipping point determines how much “room” a manager has to smooth earnings. If the
distance is large, then capability can decline significantly before the threshold is crossed. If the distance is
small, however, then efforts to offset even a small revenue shock can push the system into the downward
spiral of capability erosion. In this section we explore how the existence of the tipping point and the
degree to which the distance between the two equilibria – or the likelihood that the firm will tip into
“firefighting” – changes as a function of two key behavioral parameters: the firm’s revenue focus, , and
the degree to which the firm routinely under or overinvests in long term capability building.4
Changing revenue focus. The first parameter that influences the degree to which the system is prone to
tipping is , the level of manager’s focus on revenue vs. capability management. γ can change between 0
and 1/β (β=0.5 in our simulations), where 0≤γ≤1 reflects a capability-focused heuristic and 1<<1/ a
revenue-focused heuristic. In the previous analysis we set  to 1.5, representing moderately intense focus
4
Analytical characterization of the system is available in the online appendix S3. The capability erosion time (τ),
capability investment productivity (ρ), and total effort (h) all have a similar, more minor, impact on the results, i.e.
reducing τ.ρ.h brings the tipping point closer to the stable equilibrium, and are not discussed in detail here.
22
on smoothing earnings. In Figure 6, we show how the rate-level plot and the location of the tipping
threshold change as managers increase their propensity to focus on delivering steady earnings.
Figure 6- Sensitivity of system's dynamics to managerial response parameter γ.
As Figure 6-a suggests, strength of managers’ response to revenue shortfalls plays a significant role in
determine how robust the system is to tipping. The locations of the stable equilibria do not change since 
only affects the out-of-equilibrium behavior, but as managers respond more aggressively to short falls in
revenue (i.e. as  grows), the distance between the tipping threshold and the desired operating point
declines (see Figure 6-b), implying that a smaller shock is required to push the system over the edge.
Thus, the analysis suggests that managers who aggressively adjust short-run resource allocation to
maintain revenue in the face of demand shocks run an increased risk of pushing their firms over the
tipping threshold.
Biases in Effort Allocation. The second parameter that proves influential in determining the system’s
dynamics is the baseline allocation to revenue, u*. So far we have assumed that in equilibrium, the
managers choses the optimal allocation of effort (u*=β). The available evidence suggests, however, that
such optimality is far from guaranteed. Studies in both manufacturing and new product development
suggest that firms often over emphasize short-term revenue generation and undervalue capability
23
development (Easton and Jarrell 1998; Pfeffer and Sutton 2000; Repenning 2001; Repenning and Sterman
2002; Rahmandad 2012). These studies identify several reasons for such biases ranging from individuallevel perceptual biases—revenue generation is far more salient and tangible than long term investments
like capability development—to organizational level challenges—rewarding longer term investments in
less tangible assets like “capability” is often more difficult than recognizing contributions to immediate
revenue generation. Systematic under-investment in capability appears to often be the norm. Moreover,
rational managers who discount future rewards decrease their base-line investment in capability by a term
proportional to capability life (τ) multiplied by discount rate (r) (See Appendix S2 for details). Therefore
over-confidence in the resilience of capabilities (i.e. bias in τ estimate) and high discount rates (e.g. due to
short-term managerial incentives) can also lead to biases in u*.
Figure 7 shows how the rate level plot changes for revenue-focused managers, when the firm
over/under invests in revenue generation. When firms systematically over-invest in capability
development (here by 15%) the phase plot moves up and the distance between the desired equilibrium and
the tipping threshold increases. In this case, the system becomes more robust to temporary shocks but
this robustness comes at the expense of revenue generation since, relative to the optimum, the firm is
overinvesting in capability generation and thus receives lower revenue. Conversely, when the firm overinvests in revenue generation, the phase plot moves down and the distance between the desired
equilibrium and the tipping threshold shrinks. In fact, as we show in the figure, if the over-investment is
large enough, both the desired equilibrium and the tipping point disappear and the system displays a
single trajectory, a downward spiral to failure. Thus, as firms place greater emphasis on immediate
revenue generation, the distance between the stable equilibrium and the tipping threshold decreases and
smaller shocks are required to push the firm into the downward spiral of declining capability. In the next
section we extend this result to explore the interaction between environmental variability and a firm’s
tendency to focus on revenue.
24
Figure 7- Impact of biases in baseline effort allocation.
4.0 Extensions: Revenue Variability vs. Risk of Failure
In this section we explore how variability in underlying demand and the system’s time constants influence
the firm’s propensity to tip. To incorporate the effect of demand uncertainty we represent noise in
demand by modeling S – exogenous shocks to revenue – as a “pink” noise process, ϵ, in which normally
distributed white noise is exponentially smoothed to create first-order autocorrelation with time constant
of ξ and standard deviation σ.5
(2’)
We present large sample results in a moment, but to develop intuition for the main finding, consider two
firms responding to the same pattern of demand. The first firm focuses on maintaining revenue and
responds aggressively to any shortfall caused by a low demand realization (γ=1.8), while the second firm
focuses on maintaining capability (γ=0.2).
Two Sample Paths- Figure 8-a shows one potential sample path for demand. When the two firms face
this demand stream, the revenue-focused firm is able to maintain a steadier revenue stream (Figure 8-b).
5
. See Sterman (2000) for an extended discussion concerning the utility of pink noise as a test input for continuous
time simulation models. We truncate the noise term to stay between 0 and 2, keeping a symmetric, zero-mean, noise
term while ensuring non-negative firm output.
25
It does so by moving resources back and forth between capability development and revenue generation
more rapidly than the capability-focused firm (Figure 8-c). The steadier earnings come at the expense of
larger swings in capability level (Figure 8-d). However, as shown in panels e and f, since capability
always remains above the tipping threshold, the revenue-focused firm displays superior performance. Its
cumulative revenue over the simulation is essentially identical to that of the capability-focused
competitor6, but its revenue stream is substantially smoother. To the degree that the capital markets value
predictable earnings, they would clearly value this stability and the revenue-focused management team
would be viewed as superior. Before reaching a similar conclusion, however, consider a second
realization of demand that has identical statistical properties to the first one.
6
Deviations from optimal path lead to very slight cumulative revenue penalty for the revenue-focused
firm, but this loss is so small as not be visible in the figure.
26
Figure 8- Revenue and capability focused outcomes; one demand realization with σ=0.035; ξ=12
months.
27
Figure 9- Revenue and capability focused outcomes; a different demand realization with
σ=0.035; ξ=12 months.
Though the mean and variance are identical to the first realization, the second sample path is
characterized by extended periods of low demand between months five and twenty, and then forty and
seventy (Figure 9-a). During these periods, the revenue-focused firm works hard to maintain its revenue
(Figure 9-b) by increasingly allocating effort towards additional revenue generation ((Figure 9-c). The
strategy initially appears to be successful; during the first downturn the revenue-focused firm has higher
revenue for more than a year. However, efforts to prop-up revenue during the period of low demand
result in declining capability, which, in turn, requires more effort dedicated to revenue generation.
Eventually, the revenue-focused firm falls below the tipping threshold and the firm’s performance
declines, significantly reducing revenue and increasing variability (panels e and f). In practice such a firm
is likely to exit the market well before the end of simulated period, so the actual observations on the
outcomes of such revenue-focused firms may be right-censored. Meanwhile, the capability focused firm,
although experiencing higher revenue variability initially, survives throughout the simulation period and
generates significant more cumulative revenue.
Large Sample Results. We first consider the interaction between firms’ revenue focus, γ, and the degree
of variability in demand (σ). Figure 10 shows the results of a Monte Carlo analysis. The model was run
28
for different managerial policies (0≤γ≤2) under various levels of demand variability (0≤σ≤0.5), 1000
times for each combination of the two parameters. Contour plots report the fraction of firms that fail
(panel a), and the average recent revenue variability for the firms that survive (panel b)7. We report results
for the surviving firms (rather than the full sample) since results conditional on survival are easier to
interpret; failure significantly reduces revenue and increases its variability, confounding, in the full
sample, the day-to-day revenue outcomes with failure. Moreover, in practice firms that fail normally exit
the market, thus reporting on survivors better replicates empirically observable samples.
Figure 10-Fraction of firms that fail (a) and the average recent revenue variability of surviving
firms (b) as a function of different demand variability and managerial policies.
As discussed in the section on tipping points, capability focused firms (γ≤1) never fail thus Figure
10-a focuses on γ>1. Both higher demand variability, σ, and an increased focus on revenue are associated
with higher failure rates. Moreover, the two effects interact: for low levels of demand variability, a
revenue focus imposes only low to moderate risks. But as the environment becomes more volatile and the
distance between stable equilibrium and tipping point shrinks (See Figure 6 discussion) the chance of
failure grows significantly. In fact for volatile demand (~σ>0.1) and moderate to high focus on revenue
(~γ>1.5) the majority of firms fail.
7
Failure is defined as capability going below half the tipping value; the threshold of half is arbitrary and
results are qualitatively similar for other plausible values. Revenue variability is measured as the average
over simulation horizon of difference between revenue and its recent (12 month) exponential average.
29
Note again the challenge of learning the right lessons in this environment. Among the surviving
firms, a revenue focus reduces variability and increasingly so as the variation in demand grows (Figure
10-b). Managers who focus exclusively on capability (γ=0) pass all the variability in the environment to
their revenue, but managers who focus on revenue absorb some of that variability by changing capability
levels. At the extreme (γ=1/β) revenue shows no variability until the firm crosses the tipping point.
Consequently, among the surviving firms the average revenue shows little sensitivity to managerial policy
and demand variability. In fact, for higher values of  the surviving sample disproportionately consists of
“lucky” firms that have been exposed to long periods of better-than-average demand shocks. As a result
these lucky firms accumulate high levels of capabilities, show little variability in revenue, and experience
above average returns! The firms who are both lucky and aggressive in their efforts to manage earnings
thus provide a very tempting role-model for aspiring managers.
Interestingly, persistence of environmental shocks (ξ) and resilience of capabilities (τ) have
identical (but inverse) impacts on the system’s dynamics. The capability time constant, τ, scales the time
horizon for firm’s dynamics and ξ scales the time horizon for exogenous shocks to the system. Therefore
the system’s dynamics are indifferent to changing both parameters proportionately: the behavior of a firm
with resilient capabilities (high τ) exposed to highly correlated demand shocks (high ξ) takes more time to
unfold, but is otherwise the same as a firm with short-lived capabilities exposed to less-correlated
demand. Therefore we only analyze the large sample tradeoffs for demand variability correlation time, ξ,
in Figure 11, using graphs similar to Figure 10, keeping in mind the implications for capability resilience.
Again, tipping dynamics and failure are absent in capability focused firms (Figure 11-a). For
revenue-focused firms an increase in demand auto-correlation increases the risk of failure because periods
of bad luck are more likely to persist, pushing the revenue-focused firms over its tipping points. The
effect includes a subtle wrinkle: if more than half the firms are prone to failure due to a revenue focus,
increasing autocorrelation actually helps reduce that fraction towards 50%. In extreme, fully autocorrelated demand means 50% of the realizations consist of lucky firms with above average demand and
30
thus no tipping, and the other 50% will ultimately fail as long as γ>1. On the other hand, higher
autocorrelation reduces recent revenue variability among surviving firms as it filters out short-term
variation in demand, yet this impact is relatively small (panel b). The impact on average revenue (not
reported) is small and limited to the bias created by over-sampling “lucky” firms among survivors.
Overall, controlling for the variance in environmental shocks, firms exposed to more highly correlated
shocks are more prone to failure, especially if they are focused on revenue. Similarly, and as a corollary,
as capabilities become less resilient (a smaller τ), the chance of failure in the face of correlated shocks
also increased. Managers in environments with persistent shocks and short-lived capabilities may face a
more acute challenge in learning to balance capability investments and revenue smoothing.
Figure 11- Fraction of firms that fail (a) and the average recent revenue variability of surviving
firms (b) as a function of different demand correlation and managerial policies (σ=0.1 for
these simulations).
5.0 Discussion
Analyzing our model thus suggests that whether managing earnings helps or hurts a firm turns on the
intensity with which managers engage in it. If, as we hypothesize, capability behaves as a stock variable,
then the firm can smooth some of the demand variation it faces without risking a significant decline in
performance. Managing earnings in this situation offers several potential benefits including improved
internal financial discipline and higher external investor confidence. If, however, the firm becomes too
aggressive in its efforts to smooth earnings, it risks tipping into a vicious cycle of declining capability and
31
a growing emphasis on short-term revenue generation. Whether crossing such a tipping threshold will
result in firm failure or just an extended period of poor performance depends on both the institutional
context and the speed with which the firm realizes the need to adjust its strategy. Our model thus
reconciles the apparently puzzling tension in the literature between those who have shown that a focus on
“managing earnings” is associated with better performance in the capital markets and those who have
suggested it may be dangerously short sighted. In the space that remains we first address what is,
perhaps, the central question raised by our analysis, namely “would managers ever be so naive as to get
stuck in the trap we identify?” and then sketch an agenda for future research.
Would managers ever be sufficiently shortsighted as to allow a short-term focus to pull their firm
into the vicious cycle of capability decline highlighted in our model? Unfortunately much research on
human decision making in dynamics environments suggests that it is all too plausible for several reasons.
First, while the benefits of focusing on earnings are realized relatively quickly—cutting the travel budget
or canceling a development program can improve earnings within the given month or quarter—the
costs—poorly operating equipment or an incompetent workforce—are only realized with a significant
delay. Both experimental and field-based studies demonstrate that human decision makers do not
perform well when their actions produce multiple outcomes that occur on different time scales. People
tend to overweight those outcomes that happen quickly and to underweight those that occur with a delay
(e.g., Sterman 1989; Brehmer 1992; Repenning and Sterman 2002). Managers finding themselves in a
vicious cycle of declining capability and increasing earnings shortfalls are unlikely to make the
connection between the challenges they face and their past actions and instead are likely to redouble their
efforts to meet their short term targets (e.g., Staw 1976; Staw 1981; Azoulay, Repenning et al. 2010).
Second, because the system behaves very differently on the two sides of the tipping threshold, the
dynamics associated with tipping greatly exacerbate the challenges associated with learning to balance the
short and long run. As long as capability remains above the tipping threshold, efforts to manage earnings
will have immediate benefits and few costs; once the threshold is crossed, however the costs are
32
significant. The challenge of learning in this system is considerable. A new manager facing her first
downturn in demand might be fairly modest in her initial attempts to manage earnings, perhaps restricting
travel or temporarily cutting training budgets to offset low revenue. These efforts are likely to be
“successful” – she will meet her targets despite the downturn--thus providing powerful feedback
confirming the wisdom of her efforts. During the next down period, she is likely to be even more strict
(building on her previous success), perhaps cutting the maintenance or R&D budgets, hoping to
demonstrate further her commitment to maintaining profitability. And again, these efforts are likely to be
recognized and rewarded, providing additional positive reinforcement. This cycle of experimentation
with increasingly aggressive measures and positive reinforcement will continue until capability is pushed
below the critical threshold, whereupon performance will begin to decline despite her continued efforts.
However, having experienced several rounds of success with a strategy that cuts the resources devoted to
investing in capability, our hypothetical manager will require compelling evidence as to the error in her
ways before revisiting her strategy. Capability, in the meantime, may have fallen to a dangerously low
level.
The third factor making life difficult for managers is that it is usually difficult to measure or
reward efforts to build “capability.” Decision makers tend to focus on cues that are both salient and
tangible (Einhorn and Hogarth 1985; Einhorn and Hogarth 1986). While quarterly earnings satisfy both
of these criteria, most capabilities do not. Managers are thus unlikely to be aware that the firm’s “stock”
of capability is fluctuating dangerously. Several studies suggest, for example, that major industrial
accidents are often preceded by extended periods of inadequate investment in both training and
mechanical integrity (e.g., Baker, Gibbons et al. 2002; Columbia Accident Investigation Board 2003).
Capability in these settings had been allowed to fall precipitously, but because accidents don’t happen
every day, recognizing the costs of such a decline was difficult until the lack of capability finally
translated into a very visible accident.
33
Our analysis thus opens several avenues for further research. It suggests, for example, that firms
in settings in which capability is more easily measured, or in which the risks of tipping are more widely
understood, are less likely to sacrifice the future in favor of smoother earnings and are less likely to be
punished by the capital markets for not doing do. Most intriguingly, we believe, it reinforces the idea
that firms able to use “relational” contracts based on more “subjective” assessments to measure and
reward their managers may be able to outperform their competition. A long literature that includes
Holmstrom and Milgram’s (1991) seminal analysis of incentives in “multi-task” environments suggests
that, in the absence of relational contracts, high-powered incentives can produce significant distortions
when one activity is more easily measured than another. Given that short-term revenue is much easier to
measure objectively than capability, this dynamic is likely to reinforce the tendency of managers to focus
on current performance. If, however, managers and employees can develop an effective relational
contract (e.g., Bull 1987; Baker, Gibbons et al. 2002), they can complement the objective measures of
performance inherent in “making the numbers” with subjective measures that attempt to capture the
employee’s contribution to capability development and maintenance. Such a mechanism may, to a
degree, allow managers to “have their cake and eat it too”, to reap some of the benefits of high powered
incentives—extra effort, financial discipline, steadier earnings and better access to capital—without the
risk of tipping their firm into the vicious cycle of declining capability and increased pressure to meet
earnings. We thus suspect that a key source of above average performance might be the differential
ability of firms to create relational contracts that incorporate subjective evaluations of capability. Gibbons
and Henderson (2010) present some preliminary research exploring this idea, but it remains tantalizingly
underdeveloped.
An illustrative example of this kind of incentive scheme in action is provided by Jack Welch’s
recent description of his practice during his time as GE’s CEO (Welch and Welch 2005). While GE had a
notoriously performance-focused culture, GE’s compensation scheme under Welch’s tenure increasingly
incorporated a subjective evaluation of how each manager performed with respect to the company’s
34
“values.” Critically, GE’s definition of values extended far beyond that suggested by common usage.
While “values” typically include things like trust and integrity, Welch expanded the definition to include
those behaviors that were core to maintaining the long-term performance of the business. As he wrote,
“… (at GE) values are just behaviors—specific, nitty gritty, and so descriptive that they leave little to the
imagination (Welch and Welch 2005).” These “values” included several measures of capability that
would ultimately translate to long-term success including customer focus, product and service quality,
and the development of intellectual capital. Thus, during his tenure GE employees were evaluated both
on their ability to “hit their numbers” and on their ability to maintain and develop critical capabilities.
Further, Welch reports learning the hard way to be wary of employees that made their numbers but did
not display GE’s values. Invariably, he reports, those employees were achieving their results by
sacrificing key capabilities. “Those people,” he said, “…kill companies (Welch 2008).” While GE’s
remarkable performance during Welch’s extended tenure undoubtedly had several sources, the ability to
mix a strong performance focus with an emphasis on capability development may have been an important
factor, creating an organization capable of delivering steady performance without succumbing to the
vicious cycle of capability decline.
Conclusion
Steady and reliable earnings bring many advantages to the firms that deliver them. Share prices rise,
capital costs decline, and bonuses become both bigger and more likely. Such firms grow faster and attract
more talented people to manage that growth. Despite these benefits, both scholars and practitioners have
long been critical of the short-term focus that often characterizes western managers. Our model suggests
that the solution to this seeming paradox lies in the fundamentally non-linear nature of any enterprise
dependent on a stock of capability: above a given threshold earnings management is relatively harmless,
but below it, it can be disastrous. We hope that future research will explore both the degree to which this
model is supported empirically and the hypothesis that the ability to avoid tipping through the use of
35
relational contracts that balance attention to both short and long term investments may be a critical source
of competitive advantage.
References
Azoulay, P., N. P. Repenning, et al. (2010). "Nasty, Brutish, and Short: Embeddedness Failure in the
Pharmaceutical Industry." Administrative Science Quarterly 55(3): 472-507.
Baker, G., R. Gibbons, et al. (2002). "Relational contracts and the theory of the firm." Quarterly Journal
of Economics 117(1): 39-84.
Baker, J., F. Bowman, et al. (2007). The Report of the BP U.S. Refineries Independent Safety Review
Panel.
Barth, M. E., J. A. Elliott, et al. (1999). "Market rewards associated with patterns of increasing earnings."
Journal of Accounting Research 37(2): 387-413.
Bartov, E. (1993). "The Timing of Asset Sales and Earnings Manipulation." Accounting Review 68(4):
840-855.
Bartov, E., D. Givoly, et al. (2002). "The rewards to meeting or beating earnings expectations." Journal of
Accounting & Economics 33(2): 173-204.
Benner, M. J. (2007). "The incumbent discount: Stock market categories and response to radical
technological change." Academy of Management Review 32(3): 703-720.
Benner, M. J. (2010). "Securities Analysts and Incumbent Response to Radical Technological Change:
Evidence from Digital Photography and Internet Telephony." Organization Science 21(1): 42-62.
Bhojraj, S., P. Hribar, et al. (2009). "Making Sense of Cents: An Examination of Firms That Marginally
Miss or Beat Analyst Forecasts." Journal of Finance 64(5): 2361-2388.
Bowen, R. M., L. DuCharme, et al. (1995). "Stakeholders' implicit claims and accounting method
choice." Journal of Accounting & Economics 20(3): 255-295.
Brav, A., J. R. Graham, et al. (2005). "Payout policy in the 21st century." Journal of Financial Economics
77(3): 483-527.
Brehmer, B. (1992). "Dynamic Decision-Making - Human Control of Complex-Systems." Acta
Psychologica 81(3): 211-241.
Bull, C. (1987). "The Existence of Self-Enforcing Implicit Contracts." Quarterly Journal of Economics
102(1): 147-159.
Burgstahler, D. and I. Dichev (1997). "Earnings management to avoid earnings decreases and losses."
Journal of Accounting & Economics 24(1): 99-126.
Christensen, C. M. (1997). The innovator's dilemma : when new technologies cause great firms to fail.
Boston, Mass., Harvard Business School Press.
Columbia Accident Investigation Board (2003). Columbia Accident Investigation Report. Burlington,
Ontario, Apogee Books.
Dallas, L. (2011). "Short-Termism, the Financial Crisis, and Corporate Governance." Journal of
Corporation Law 37: 264.
Dierickx, I. and K. Cool (1989). "Asset Stock Accumulation and Sustainability of Competitive
Advantage." Management Science 35(12): 1504-1511.
Drucker, P. (1986). A crisis of capitalism. Wall Street Journal. (September 30): 31.
Dustmann, C., B. Fitzenberger, et al. (2014). "From Sick Man of Europe to Economic Superstar:
Germany's Resurgent Economy." Journal of Economic Perspectives 28(1): 167-188.
36
Easton, G. S. and S. L. Jarrell (1998). "The effects of total quality management on corporate performance:
An empirical investigation." Journal of Business 71(2): 253-307.
Einhorn, H. J. and R. M. Hogarth (1985). "Ambiguity and Uncertainty in Probabilistic Inference."
Psychological Review 92(4): 433-461.
Einhorn, H. J. and R. M. Hogarth (1986). "Judging Probable Cause." Psychological Bulletin 99(1): 3-19.
Farrell, K. A. and D. A. Whidbee (2003). "Impact of firm performance expectations on CEO turnover and
replacement decisions." Journal of Accounting & Economics 36(1-3): 165-196.
Gibbons, R. and R. Henderson (2010). What do managers do? Exploring Persistent Performance
Differences among Seemingly Similar Enterprises. The Handbook of Organizational Economics.
R. Gibbons and J. Roberts. Princeton and Oxford, Princeton University Press: 680-731.
Graham, J. R., C. R. Harvey, et al. (2005). "The economic implications of corporate financial reporting."
Journal of Accounting & Economics 40(1-3): 3-73.
Gunny, K. A. (2010). "The Relation Between Earnings Management Using Real Activities Manipulation
and Future Performance: Evidence from Meeting Earnings Benchmarks." Contemporary
Accounting Research 27(3): 855-+.
Haldane, A. G. (2010). Patience and Finance. Oxford China Business Forum. Beijing.
Healy, P. M. and J. M. Wahlen (1999). "A review of the earnings management literature and its
implications for standard setting." Accounting Horizons 13(4): 365-383.
Helfat, C. E. (2007). Dynamic capabilities : understanding strategic change in organizations. Malden,
MA, Blackwell Pub.
Holmstrom, B. and P. Milgrom (1991). "Multitask Principal Agent Analyses - Incentive Contracts, Asset
Ownership, and Job Design." Journal of Law Economics & Organization 7: 24-52.
Ingrassia, P. (2010). Crash Course: The American Automobile Industry’s Road to Bankruptcy and
Bailout—and Beyond. New York, Random House.
Laverty, K. J. (1996). "Economic ''short-termism'': The debate, the unresolved issues, and the implications
for management practice and research." Academy of Management Review 21(3): 825-860.
Peek, J. and E. S. Rosengren (2005). "Unnatural selection: Perverse incentives and the misallocation of
credit in Japan." American Economic Review 95(4): 1144-1166.
Pfeffer, J. and R. I. Sutton (2000). The knowing-doing gap : how smart companies turn knowledge into
action. Boston, Mass., Harvard Business School Press.
Rahmandad, H. (2012). "Impact of Growth Opportunities and Competition on Firm-Level Capability
Development Trade-offs." Organization Science 23(1): 138-154.
Repenning, N. P. (2001). "Understanding fire fighting in new product development." The Journal of
Product Innovation Management 18: 285-300.
Repenning, N. P. and J. D. Sterman (2001). "Nobody ever gets credit for fixing problems that never
happened: Creating and sustaining process improvement." California Management Review 43(4):
64-+.
Repenning, N. P. and J. D. Sterman (2002). "Capability traps and self-confirming attribution errors in the
dynamics of process improvement." Administrative Science Quarterly 47(2): 265-295.
Rudolph, J. W. and N. P. Repenning (2002). "Disaster Dynamics: Undrstanding the Role of Quantity in
Organizational Collapse." Administrative Science Quarterly 47: 1-30.
Rummel, N. (2008). Short-term earnings focus led to Wall Street woes, former SEC chief Donaldson
says. Financial Week. Available at
http://www.financialweek.com/apps/pbcs.dll/article?AID=/20080917/REG/809179981/1038/EX
CLUSIVES.
37
Staw, B. M. (1976). "Knee-deep in the Big Muddy: A study of escalating commitment to a chosen course
of action." Organizational Behavior and Human Performance 16: 27-44.
Staw, B. M. (1981). "The escalation of commitment to a course of action." Academy of Management
Review 6: 577-587.
Stein, J. C. (1989). "Efficient Capital-Markets, Inefficient Firms - a Model of Myopic CorporateBehavior." Quarterly Journal of Economics 104(4): 655-669.
Sterman, J. (2000). Business dynamics: systems thinking and modeling for a complex world. Boston,
McGraw-Hill/Irwin.
Sterman, J. D. (1989). "Modeling Managerial Behavior - Misperceptions of Feedback in a Dynamic
Decision-Making Experiment." Management Science 35(3): 321-339.
Sterman, J. D. (1994). "Learning in and about complex systems." System Dynamics Review 10(2-3): 91330.
Strogatz, S. H. (1994). Nonlinear dynamics and Chaos : with applications to physics, biology, chemistry,
and engineering. Reading, Mass., Addison-Wesley Pub.
The Aspen Institute (2009). Overcoming Short-termism: A Call for a More Responsible Approach to
Investment and Business Management. September 9, 2009, The Aspen Institute Business and
Society Program.
Welch, J. (2008). Unpublished Personal Interview.
Welch, J. and S. Welch (2005). Winning. New York, HarperBusiness Publishers.
Williams, H. (1980). The Economy and the Future—The Tyranny of the Short-Run. An address given to
the Commonwealth Club of San Francisco. 21 November. Available at
http://www.sec.gov/news/speech/1980/112180williams.pdf.
38
Online Appendix to Accompany the Paper:
MAKING THE NUMBERS?
“SHORT TERMISM” & THE PUZZLE OF ONLY OCCASIONAL DISASTER
S1- The finite horizon optimal allocation
The optimal control problem for maximizing expected revenue in our problem can be written as
maximizing the expected revenue subject to the dynamics of the system and the budget constraint:
∫
Subject to:
To solve this problem we can set up the present value Hamiltonian with the co-state variable λ
which represents the shadow price of the capability at any point in time. Noting that environmental
shock to the revenue, S, has a mean of zero and is not correlated with the other component of
revenue and is zero-mean, we have
, and thus we find a rather simple Hamiltonian
function:
(
)
The necessary conditions for finding the optimal allocation policy, u, is:
39
These conditions are also sufficient because the Hamiltonian is concave with respect to u and C for
feasible values of C and u. Solving the first constraint we find the following optimal allocation
fraction:
( (
)
)
After replacing the optimum allocation in the second condition, the dynamics of co-state variable is
described by the following differential equation:
(
)
Solving this differential equation (using Bernoulli method) we get the following time trajectory for
the shadow price of capability, λ:
(
)
(
)
Using the end state condition (λ(T)=0) we can solve for the constant K which gives the analytical
expression for the time trajectory of λ for any time horizon and combination of parameters.
(
)
Inspecting the results, we note that this constant term is negative, and very small as long as T>τ(1β), that is, the time to end of horizon is appreciably smaller than the time constant for the erosion of
capability. Therefore λ is almost constant until we get fairly close (relative to τ) to the end of
40
investment horizon (T), at which time the shadow price starts to decline precipitously (note the
exponential term in equation for λ), leading to increasing allocation of resources to revenue
generation and a decline of capability, until at exactly time T the shadow price and capability stocks
both become zero.
Therefore, assuming T>τ(1-β), we can find a constant shadow price of capability that applies for a
large section of our time horizon:
(
(
)
)
Replacing λ with this steady state value in equation for uDyn and simplifying the equations we get the
following expression for the approximate optimal control allocation:
(
)
This simple expression suggests that optimal allocation in the dynamics case is 1) consistent with
the steady state allocation, that is, if capability is at the steady-state optimal level, the allocation will
be the same as steady state. 2) Variations of the optimal path in capability are compensated for by
linear shifts in the allocation fraction: When capability falls short of the optimal steady state value,
the allocation favors capability investment, while too much capability (relative to steady state) will
lead to more effort (than steady state optimal) being allocated to revenue generation. Note that the
heuristic used in our paper simplifies to this function when γ =0.
S2- The infinite horizon optimal allocation
The infinite horizon optimal control problem with discounted revenue can be written as:
41
∫
Subject to:
Here r is the continuous time discount rate. To solve this problem we set up the current value
Hamiltonian with the transformed co-state variable
and follow the regular steps:
(
)
To find the optimal allocation policy, u, we solve the following equations:
Solving the first constraint we find the following optimal allocation fraction which is similar to the
finite horizon case:
( (
)
)
After replacing the optimum allocation in the second condition, the dynamics of co-state variable is
described by the following equation:
(
)
42
In this case we observe that the equilibrium
value satisfies the terminal conditions, and thus
provides the following solutions for the optimal co-state trajectory and allocation:
(
(
)
(
)
)
In the infinite horizon discounted case the optimal allocation differs from the finite horizon, undiscounted,
case with a factor of (1+rτ): if capabilities are slow to erode (large τ) and if discount rate is high, the
baseline allocation favors revenue generation beyond the steady state optimal allocation. Moreover the
infinite horizon case does not include the precipitous decline in the value of capability at the end of time
horizon (because there is no end to the time horizon).
S3- The effort allocation function and characteristic of the resulting phase diagram
Variables and model definition (reproduced from the paper)
Revenue Function:
Allocated effort to revenue
Allocated effort to capability
System’s dynamics
Optimal steady state allocation policy
Allocation heuristic used in this study
(
(
) ( )
)
Target revenue
43
Expected revenue using optimal steady state policy
Effort to revenue under optimal steady state policy
Capability using optimal steady state policy
Throughout the rest of the document it is assumed that we are using a constant return to scale
production function (α+β=1).
Phase diagram
The phase diagram for the system reflects the changes in capability (dC/dt) as a function of
capability. Specifically, replacing the equation for allocation into the system’s dynamics, we get:
(
(
(
) (
)
))
after replacement and simplification, we get:
̇
(
(
(
)
))
The allocation function thus has one adjustment factor that responds to the capability level, and
another that responds to environmental shocks. The latter component reduces variability in
response to the environmental shocks and the former either smoothes revenue (γ>1) or fixes
capability shortfalls (γ<1) in response to deviations of capability. The response to capability level is
thus the result of two competing forces, one which attempts to align the capability level with the
optimal trajectory based on the optimal control policy, and another which compensates for falling
capability by increasing allocation to revenue generation, thus smoothing the revenue trajectory.
These forces are at balance when γ=1, capability renewal tendencies win for smaller γ and revenue
smoothing dominates for
.
For simplifying the analysis of the system, we focus on the deterministic version of the equation,
where the impact of environmental noise is excluded from calculations of capability change.:
44
̇
(
(
(
)
))
By equating this equation to zero we find that it always has a fixed point at
where capability
equals the optimal steady state capability. The existence and number of other fixed points depends
on γ.
1)
:
For
we have:
̇
(
(
)
)
Yet the right hand side of inequality is by definition zero, so ̇
Using a similar argument, it is easy to see that for
always negative. Therefore for
.
the net rate of change in capability is
the phase diagram includes a single equilibrium at
and no
other fixed points, the system will always move back towards this equilibrium.
2)
:
Calling
( )
, the net flow equation has two ranges:
̇
̇
(
(
)
)
Therefore at least one additional fixed point exists at C=0. We now focus on the behavior of ̇ when
.
First, we observe that ̇ is a continuous function in C that at
takes the
value and at
is zero. The
extremum for ̇ can be found by equating its derivative with respect to C to zero which provides the
following unique solution:
45
̇
(
)
On the other hand:
̇
All the terms in the equation are positive, except for the one negative sign, therefore the second derivative
of capability flow with respect to capability is always negative in this region. As a result the extremum
found above is the only maximum for the net capability flow function which should be above zero (given
that ̇
at
) and thus there is a single other point at which
̇
. This point can be found
numerically8 by solving the rate-level equation for zero capability change rate. Given the positive first
derivative of ̇ with respect to C at this point, it is also the only tipping point for the system.
To recap, when
, the system includes a single unique equilibrium at
includes exactly three fixed points: two are stable equilibria at
. For
the system
and one is a tipping point
located in between.
8
No general analytical solution exists for the location of tipping point.
46